Re: False negatives with compare?

On 30 December 2009 pm 11:07:18 
da.mihi.sis.bubulae.frustrum.assae.solana.tuberosa.in.modo.gallico.fricta.ac.quassum.lactatum.coagulatum.crassum@imagemagick.org 
wrote:
> > but as far as I can tell both images are identical.
> 
> One has an alpha channel, the other does not:
> 
>   -> compare +matte -metric rmse external.png  external-crushed.png null:
>   0 (0)


Thanks, that worked for these 2 images.

But hmm, doesn't adding the +matte switch makes the comparison less generic? I 
do have a few pngs with transparency enabled, for which I would need to 
compare both rgb data and alpha data to ensure everything is still identical. 
Is there a set of arguments I could use to do a generic comparison in all 
cases?

Reading on the -channel switch, it seems I should turn "-alpha on" which 
creates an opaque alpha channel if its not in an image, but that didn't work 
for these 2 images :(. "-alpha set" did work though.

For these 2 images anyway, I don't understand how the alpha channel is set in 
the first image. I thought that even though there is alpha data, it was fully 
opaque. In gimp if I push a plain red layer underneath for example, there's no 
color change so the icon is fully opaque, which is why I guess the png crusher 
(pngout) dropped it in the optimized image.

Sorry I'm kind of new to this, how would I go about checking out the actual 
structure of the first non-optimized image?

Thanks!
Tim.